Chapter 2 HPSS PlanningHPSS Installation Guide September 2002 95Release 4.5, Revision 2cannot be greater than half the number of drives available. Also, doing multiple copies from diskto two tape storage classes with the same media type will perform very poorly if the stripe widthin either class is greater than half the number of drives available. The recover utility also requires anumber of drives equivalent to 2 times the stripe width to be available to recover data from adamaged virtual volume if invoked with the repack option.Guideline 2: Select a stripe width that results in data transmission rates from the drives matchingor being less than what is available through rates from the network.Explanation: Having data transmission off the devices that is faster than the network will wastedevice resources, since more hardware and memory (for Mover data buffers) will be allocated tothe transfer, without achieving any performance improvement over a smaller stripe width. Also, ifa large number of concurrent transfers are frequently expected, it may be better (from an overallsystem throughput point of view) to use stripe widths that provide something less than thethroughput supported by the network - as the aggregate throughput of multiple concurrentrequests will saturate the network and overall throughput will be improved by requiring lessdevice and memory resources.Guideline 3: For smaller files, use a small stripe width or no striping at all.Explanation: For tape, the situation is complex. If writing to tape directly, rather than via diskmigration, writing a file will usually result in all the tape volumes having to be mounted andpositioned before data transmission can begin. This latency will be driven by how many mountscan be done in parallel, plus the mount time for each physical volume. If the file being transmittedis small, all of this latency could cause performance to be worse than if a smaller stripe or nostriping were used at all.As an example of how to determine stripe width based on file size and drive performance, imaginea tape drive that can transmit data at about 10 MB/second and it takes about 20 seconds on averageto mount and position a tape. For a one-way stripe, the time to transmit a file would be: / 10 + 20Now consider a 2-way stripe for this storage class which has only one robot. Also assume that thisrobot has no capability to do parallel mounts. In this case, the transmission time would be: / 20 + 2 * 20An algebraic calculation indicates that the single stripe would generally perform better for files thatare less than 400 MB in size.Guideline 4: Migration can use larger stripe widths.Explanation: For migration operations from disk, the tape virtual volume usually is mounted andpositioned only once. In this case, larger stripe widths can perform much better than smaller. Thenumber of drives available for media in this storage class also should be a multiple of the stripewidth. If not, less than optimal use of the drives is likely unless the drives are shared across storageclasses.
